Meta has publicly launched Muse Spark 1.1, a multimodal AI model built for agentic coding tasks. The release puts Meta in direct competition with OpenAI and Anthropic, both of which have had similar products on the market for some time. This is Meta playing catch-up, but that does not make it a non-factor.
The model can handle multistep reasoning, manage digital workflows, fix bugs, and help enterprises run large code migrations. Meta's pitch is straightforward: Spark can take on the kind of complex, multi-stage automation work that big companies are increasingly willing to pay AI providers to handle. Mark Zuckerberg, who had not posted on X since July 2023, broke his three-year silence on the platform to call Spark “a strong agentic and coding model at a very low price.”
The launch is part of a busy week for Meta in AI. The company also released Muse Image, an AI image-generation model, earlier in the week. And it is not just Meta making noise right now. OpenAI dropped its GPT-5.6 model family on the same day, and SpaceXAI released a new version of Grok. The AI market is crowded, and every company in it is fighting for the same enterprise budgets.
Pricing is where Meta may have the clearest shot at winning customers. According to Reuters, Spark 1.1 will cost $1.25 per million input tokens and $4.25 per million output tokens. That puts it roughly in line with Anthropic's Claude Haiku 4.5 and OpenAI's GPT-5.6 Luna, though slightly above both. For enterprise buyers comparing costs across multiple providers, even small differences in token pricing matter at scale.
Meta says Spark is strongest in three areas:
- Agentic performance across external apps and services
- Tool use and multi-step task execution
- Computer use, meaning the model can interact with software interfaces directly
These are exactly the capabilities enterprises want when they think about automating developer workflows. Code migration, in particular, is a painful and expensive process for large organizations, and AI tools that can handle it reliably have a real market. Meta is betting that Spark is good enough at these tasks to pull customers away from more established options.
The original version of Spark was announced in April. The 1.1 release is the first public launch. Zuckerberg's note that there is “more to come soon” suggests Meta is planning to move quickly with follow-up releases, which would be consistent with the pace most major AI labs are keeping right now. Falling behind in model releases, even briefly, can mean losing developer mindshare to competitors who ship faster.
Whether Meta can close the gap with OpenAI and Anthropic in the coding AI space will depend on more than pricing. Developers tend to stick with tools they trust, and trust is built through consistent performance over time. Meta is starting that process now, later than its rivals, but with the scale and resources to make a serious run at the market.
